For a document image obtained through scanning or photographing, e.g. a document image scanned by an overhead scanner, since the image is illuminated by for example a light source (such as LED) on the scanner, the scanned image obtained possibly has a shadow. Due to the existence of the shadow, the quality of the image is deteriorated. Therefore, it is necessary to remove the shadow such that a background of the image is uniform.
The currently-used method of removing a shadow is to adjust an image intensity of a document image according to a brightness image obtained by estimating from the document image, so as to remove the shadow, thereby obtaining a document image with a uniform background. Hereinafter, the fundamental principle of removing a shadow will be described in combination with equations (1) and (2). As shown in equation (1):IY=LY·RY  (1)
where IY represents an input image, LY represents a brightness image, and RY represents a reflected image.
Therefore, an image with a uniform background can be obtained from equation (2):
                                          I            ~                    Y                =                              c            ⁢                                                  ⁢                          R              Y                                =                                    cI              Y                                                      L                Y                            +              ɛ                                                          (        2        )            
where c is a constant, which means an average image gray scale; and ε is a very small constant, serving to prevent the denominator from being 0.
FIG. 1 is a view illustrating a document image obtained through scanning ((a) in FIG. 1), a brightness image estimated from document image ((b) in FIG. 1) and a document image with a uniform background after removal of a shadow ((c) in FIG. 1). It follows that obtaining an accurate brightness image is the key for removing a shadow from a document image. At present, more and more attention is being drawn to the problem as to how to obtain a more accurate brightness image.